System, method, and recording medium for knowledge graph augmentation through schema extension

Abstract

A method, system, and recording medium for knowledge graph augmentation using data based on a statistical analysis of attributes in the data, including mapping classes, attributes, and instances of the classes of the data, indexing semantically similar input data elements based on the mapped data using at least one of a label-based analysis, a content-based analysis, and an attribute-based clustering, and ranking the semantically similar input data elements to create a ranked list.

What is claimed is: 
     
       1. A method for knowledge graph augmentation, comprising:
 mapping:
 classes from an input of structured data including columns and rows to data in separate structured data including columns and rows; 
 attributes in the separate structured data corresponding to the classes in the input of structured data; and 
 instances of the classes of the data in the separate structured data and a data string corresponding to the instances; 
 
 indexing semantically similar input data elements corresponding to the attributes for the data string corresponding to the instances of the classes of data by performing computations with the mapped data using at least one of a label-based analysis, a content-based analysis, and an attribute-based clustering as a basis for the computations; and 
 ranking the semantically similar input data elements to create a ranked list of attributes to augment the input of structured data and populate with the data string corresponding to the instances, 
 wherein the ranking combines a set of filters to refine the ranked list of attributes, the set of filters including:
 a first filter according to column ranges of the columns; 
 a second filter according to a column uniqueness of the columns; 
 a third filter according to a type of data in a column of the columns; 
 a fourth filter according to a distribution of values in the columns. 
 
 
     
     
       2. The method of  claim 1 , further comprising clustering, via a graph clustering algorithm, the semantically similar data elements into sets of attributes with a same semantic intention. 
     
     
       3. The method of  claim 1 , wherein the ranking comprises a coherence ranking of the attributes of the semantically similar input data elements. 
     
     
       4. The method of  claim 3 , wherein the ranking further comprises ranking, via a page-ranking analysis, the semantically similar input data elements. 
     
     
       5. The method of  claim 1 , wherein the ranking comprises a consistency ranking of the attributes of the semantically similar input data elements. 
     
     
       6. The method of  claim 5 , wherein the ranking further comprises ranking, via a page-ranking analysis, the semantically similar input data elements. 
     
     
       7. The method of  claim 1 , wherein the ranking comprises a class-based ranking of the attributes of the semantically similar input data elements. 
     
     
       8. The method of  claim 1 , wherein the ranking comprises at least two of a coherence ranking of the attributes, a consistency ranking of the attributes, and a class-based ranking of the attributes, thereby to create a plurality of ranked lists of the semantically similar input data elements. 
     
     
       9. The method of  claim 1 , wherein the ranking comprises a coherence ranking of the attributes, a consistency ranking of the attributes, and a class-based ranking of the attributes, to create a plurality of ranked lists. 
     
     
       10. The method of  claim 9 , further comprising merging the plurality of ranked lists into a single ranked list. 
     
     
       11. The method of  claim 1 , wherein the indexing indexes the semantically similar input data elements by performing computations with at least two of the label-based analysis, a content-based analysis, and an attribute-based clustering as a basis for the computations. 
     
     
       12. The method of  claim 1 , wherein the indexing indexes the semantically similar input data elements by performing computations with a label-based analysis, a content-based analysis, and an attribute-based clustering as a basis for the computations. 
     
     
       13. The method of  claim 1 , wherein the ranking further comprises ranking, via a page-ranking analysis, the semantically similar input data elements. 
     
     
       14. The method of  claim 1 , wherein the ranking further comprises ranking the semantically similar input data elements by performing computations via a page-ranking analysis combined with at least one of a coherence ranking of the attributes, a consistency ranking of the attributes, and a class-based ranking of the attributes as a basis for the computations via the page-ranking analysis. 
     
     
       15. The method of  claim 1 , wherein the ranking further comprises ranking, via a page-ranking analysis, the semantically similar input data elements,
 wherein the page-ranking analysis is based on a publication date of the page. 
 
     
     
       16. The method of  claim 1 , wherein the ranking further comprises ranking, via a page-ranking analysis, the semantically similar input data elements combined with each of a coherence ranking of the attributes, a consistency ranking of the attributes, and a class-based ranking of the attributes. 
     
     
       17. The method of  claim 1 , wherein the ranking comprises each of:
 a coherence ranking of the attributes of the semantically similar input data elements; 
 a consistency ranking of the attributes of the semantically similar input data elements; and 
 a class-based ranking of the attributes of the semantically similar input data elements, and 
 wherein a maximum measure of trust for a given cluster is used for the ranked list from the coherence ranking, consistency ranking, or the class-based ranking. 
 
     
     
       18. A non-transitory computer-readable recording medium recording a knowledge graph augmentation program, the program causing a computer to perform:
 mapping:
 classes from an input of structured data to data in separate structured data; 
 attributes in the separate structured data corresponding to the classes in the input of structured data; and 
 instances of the classes of the data in the separate structured data and a data string corresponding to the instances; 
 
 indexing semantically similar input data elements corresponding to the attributes for the data string corresponding to the instances of the classes of data by performing computations with the mapped data using at least one of a label-based analysis, a content-based analysis, and an attribute-based clustering as a basis for the computations; and 
 ranking the semantically similar input data elements to create a ranked list of attributes to augment the input of structured data and populate with the data string corresponding to the instances, 
 wherein the ranking combines a set of filters to refine the ranked list of attributes, the set of filters including:
 a first filter according to column ranges of the columns; 
 a second filter according to a column uniqueness of the columns; 
 a third filter according to a type of data in a column of the columns; 
 a fourth filter according to a distribution of values in the columns. 
 
 
     
     
       19. A system for knowledge graph augmentation, comprising:
 a mapping device configured to map: classes from an input of structured data to data in separate structured data;
 attributes in the separate structured data corresponding to the classes in the input of structured data; and 
 instances of the classes of the data in the separate structured data and a data string corresponding to the instances; 
 
 an indexing device configured to index semantically similar input data elements corresponding to the attributes for the data string corresponding to the instances of the classes of data by performing computations with the mapped data using at least one of a label-based analysis, a content-based analysis, and an attribute-based clustering as a basis for the computations; and 
 a ranking device configured to rank the semantically similar input data elements to create a ranked list of attributes to augment the input of structured data and populate with the data string corresponding to the instances, 
 wherein the ranking device is further configured to combine a set of filters to refine the ranked list of attributes, the set of filters including:
 a first filter according to column ranges of the columns; 
 a second filter according to a column uniqueness of the columns; 
 a third filter according to a type of data in a column of the columns; 
 a fourth filter according to a distribution of values in the columns.